Lexer.x 55.1 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
--
-- GHC's lexer.
--
-- This is a combination of an Alex-generated lexer from a regex
-- definition, with some hand-coded bits.
--
-- Completely accurate information about token-spans within the source
-- file is maintained.  Every token has a start and end SrcLoc attached to it.
--
-----------------------------------------------------------------------------

--   ToDo / known bugs:
--    - Unicode
--    - parsing integers is a bit slow
--    - readRational is a bit slow
--
--   Known bugs, that were also in the previous version:
--    - M... should be 3 tokens, not 1.
--    - pragma-end should be only valid in a pragma

{
module Lexer (
25
   Token(..), lexer, pragState, mkPState, PState(..),
26
   P(..), ParseResult(..), getSrcLoc, 
27
   failLocMsgP, failSpanMsgP, srcParseFail,
28
   getMessages,
29
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
30
   getLexState, popLexState, pushLexState,
31
   extension, standaloneDerivingEnabled, bangPatEnabled
32 33 34 35
  ) where

#include "HsVersions.h"

36 37
import Bag
import ErrUtils
38 39 40 41 42 43
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
44
import DynFlags
45
import Ctype
46
import Util		( maybePrefixMatch, readRational )
47

48
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
49
import Data.Bits
50
import Data.Char 	( chr, isSpace )
Simon Marlow's avatar
Simon Marlow committed
51 52
import Data.Ratio
import Debug.Trace
53 54

#if __GLASGOW_HASKELL__ >= 605
55
import Data.Char 	( GeneralCategory(..), generalCategory, isPrint, isUpper )
56
#else
57
import Compat.Unicode	( GeneralCategory(..), generalCategory, isPrint, isUpper )
58
#endif
59 60
}

61
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
62
$whitechar   = [\ \n\r\f\v\xa0 $unispace]
63
$white_no_nl = $whitechar # \n
64
$tab         = \t
65 66

$ascdigit  = 0-9
67
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
68
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
69 70 71
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
72
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~ \xa1-\xbf \xd7 \xf7]
73
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
74 75
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

76
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
77 78 79
$asclarge  = [A-Z \xc0-\xd6 \xd8-\xde]
$large     = [$asclarge $unilarge]

80
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
81 82 83
$ascsmall  = [a-z \xdf-\xf6 \xf8-\xff]
$small     = [$ascsmall $unismall \_]

84
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
85
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
86 87

$octit	   = 0-7
88
$hexit     = [$decdigit A-F a-f]
89 90 91 92
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

93 94
$docsym    = [\| \^ \* \$]

95 96 97 98 99 100
@varid     = $small $idchar*
@conid     = $large $idchar*

@varsym    = $symbol $symchar*
@consym    = \: $symchar*

101
@decimal     = $decdigit+
102 103 104 105 106 107 108 109 110
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

-- we support the hierarchical module name extension:
@qual = (@conid \.)+

@floating_point = @decimal \. @decimal @exponent? | @decimal @exponent

111 112 113 114 115
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

116 117 118 119
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
120
$tab+         { warn Opt_WarnTabs (text "Tab character") }
121 122 123 124 125

-- Everywhere: deal with nested comments.  We explicitly rule out
-- pragmas, "{-#", so that we don't accidentally treat them as comments.
-- (this can happen even though pragmas will normally take precedence due to
-- longest-match, because pragmas aren't valid in every state, but comments
126 127 128 129
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
130 131 132 133

-- Single-line comments are a bit tricky.  Haskell 98 says that two or
-- more dashes followed by a symbol should be parsed as a varsym, so we
-- have to exclude those.
134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167

-- Since Haddock comments aren't valid in every state, we need to rule them
-- out here.  

-- The following two rules match comments that begin with two dashes, but
-- continue with a different character. The rules test that this character
-- is not a symbol (in which case we'd have a varsym), and that it's not a
-- space followed by a Haddock comment symbol (docsym) (in which case we'd
-- have a Haddock comment). The rules then munch the rest of the line.

"-- " ~$docsym .* ;
"--" [^$symbol : \ ] .* ;

-- Next, match Haddock comments if no -haddock flag

"-- " $docsym .* / { ifExtension (not . haddockEnabled) } ;

-- Now, when we've matched comments that begin with 2 dashes and continue
-- with a different character, we need to match comments that begin with three
-- or more dashes (which clearly can't be Haddock comments). We only need to
-- make sure that the first non-dash character isn't a symbol, and munch the
-- rest of the line.

"---"\-* [^$symbol :] .* ;

-- Since the previous rules all match dashes followed by at least one
-- character, we also need to match a whole line filled with just dashes.

"--"\-* / { atEOL } ;

-- We need this rule since none of the other single line comment rules
-- actually match this case.

"-- " / { atEOL } ;
168 169 170 171 172 173 174 175 176 177 178 179 180

-- 'bol' state: beginning of a line.  Slurp up all the whitespace (including
-- blank lines) until we find a non-whitespace character, then do layout
-- processing.
--
-- One slight wibble here: what if the line begins with {-#? In
-- theory, we have to lex the pragma to see if it's one we recognise,
-- and if it is, then we backtrack and do_bol, otherwise we treat it
-- as a nested comment.  We don't bother with this: if the line begins
-- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
<bol> {
  \n					;
  ^\# (line)?				{ begin line_prag1 }
181
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
182
  ^\# \! .* \n				; -- #!, for scripts
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
  ()					{ do_bol }
}

-- after a layout keyword (let, where, do, of), we begin a new layout
-- context if the curly brace is missing.
-- Careful! This stuff is quite delicate.
<layout, layout_do> {
  \{ / { notFollowedBy '-' }		{ pop_and open_brace }
	-- we might encounter {-# here, but {- has been handled already
  \n					;
  ^\# (line)?				{ begin line_prag1 }
}

-- do is treated in a subtly different way, see new_layout_context
<layout>    ()				{ new_layout_context True }
<layout_do> ()				{ new_layout_context False }

-- after a new layout context which was found to be to the left of the
-- previous context, we have generated a '{' token, and we now need to
-- generate a matching '}' token.
<layout_left>  ()			{ do_layout_left }

205
<0,option_prags> \n				{ begin bol }
206 207 208 209 210

"{-#" $whitechar* (line|LINE) 		{ begin line_prag2 }

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
211
<line_prag1> $decdigit+			{ setLine line_prag1a }
212
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
213 214 215 216
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
217
<line_prag2> $decdigit+			{ setLine line_prag2a }
218
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
219 220 221
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
222

223 224 225
-- We only want RULES pragmas to be picked up when explicit forall
-- syntax is enabled is on, because the contents of the pragma always
-- uses it. If it's not on then we're sure to get a parse error.
226
-- (ToDo: we should really emit a warning when ignoring pragmas)
227 228
-- XXX Now that we can enable this without the -fglasgow-exts hammer,
-- is it better just to let the parse error happen?
229
<0>
230
  "{-#" $whitechar* (RULES|rules) / { ifExtension explicitForallEnabled } { token ITrules_prag }
231

232
<0,option_prags> {
233 234 235 236 237
  "{-#" $whitechar* (INLINE|inline)	{ token (ITinline_prag True) }
  "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
  					{ token (ITinline_prag False) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
  					{ token ITspec_prag }
238
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
239 240 241 242
	$whitechar* (INLINE|inline)	{ token (ITspec_inline_prag True) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
	$whitechar* (NO(T?)INLINE|no(t?)inline)
					{ token (ITspec_inline_prag False) }
243 244 245 246
  "{-#" $whitechar* (SOURCE|source)	{ token ITsource_prag }
  "{-#" $whitechar* (DEPRECATED|deprecated)
  					{ token ITdeprecated_prag }
  "{-#" $whitechar* (SCC|scc)		{ token ITscc_prag }
andy@galois.com's avatar
andy@galois.com committed
247 248
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
249
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
250
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
251

252 253 254 255
  "{-#" $whitechar* (DOCOPTIONS|docoptions)
  / { ifExtension haddockEnabled }     { lex_string_prag ITdocOptions }

 "{-#"                                 { nested_comment lexToken }
256 257 258 259 260

  -- ToDo: should only be valid inside a pragma:
  "#-}" 				{ token ITclose_prag}
}

261 262 263 264 265 266 267
<option_prags> {
  "{-#" $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#" $whitechar* (OPTIONS_GHC|options_ghc)
                                        { lex_string_prag IToptions_prag }
  "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
  "{-#" $whitechar* (INCLUDE|include)   { lex_string_prag ITinclude_prag }
}
268

269
<0,option_prags> {
270
	-- This is to catch things like {-# OPTIONS OPTIONS_HUGS ... 
271
  "{-#" $whitechar* $idchar+		{ nested_comment lexToken }
272 273
}

274 275
-- '0' state: ordinary lexemes

276 277
-- Haddock comments

278
<0> {
279 280
  "-- " $docsym    / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
281 282
}

283 284
-- "special" symbols

285
<0> {
286 287 288 289
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
290
<0> {
291 292 293 294 295 296 297 298 299 300
  "[|"	    / { ifExtension thEnabled }	{ token ITopenExpQuote }
  "[e|"	    / { ifExtension thEnabled }	{ token ITopenExpQuote }
  "[p|"	    / { ifExtension thEnabled }	{ token ITopenPatQuote }
  "[d|"	    / { ifExtension thEnabled }	{ layout_token ITopenDecQuote }
  "[t|"	    / { ifExtension thEnabled }	{ token ITopenTypQuote }
  "|]"	    / { ifExtension thEnabled }	{ token ITcloseQuote }
  \$ @varid / { ifExtension thEnabled }	{ skip_one_varid ITidEscape }
  "$("	    / { ifExtension thEnabled }	{ token ITparenEscape }
}

301
<0> {
ross's avatar
ross committed
302 303
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
304 305 306
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

307
<0> {
308 309 310
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

311
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
312 313 314 315 316 317
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

318
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
319 320
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
321 322
}

323
<0,option_prags> {
324 325 326 327 328 329 330 331 332 333 334 335
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

336
<0,option_prags> {
337 338 339 340 341 342 343 344 345 346 347 348 349
  @qual @varid			{ check_qvarid }
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

-- after an illegal qvarid, such as 'M.let', 
-- we back up and try again in the bad_qvarid state:
<bad_qvarid> {
  @conid			{ pop_and (idtoken conid) }
  @qual @conid			{ pop_and (idtoken qconid) }
}

350
<0> {
351 352 353 354
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
355 356 357 358
}

-- ToDo: M.(,,,)

359
<0> {
360 361 362 363 364 365
  @qual @varsym			{ idtoken qvarsym }
  @qual @consym			{ idtoken qconsym }
  @varsym			{ varsym }
  @consym			{ consym }
}

366 367
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
368
<0> {
369 370 371 372 373 374 375
  -- Normal integral literals (:: Num a => a, from Integer)
  @decimal			{ tok_num positive 0 0 decimal }
  0[oO] @octal			{ tok_num positive 2 2 octal }
  0[xX] @hexadecimal		{ tok_num positive 2 2 hexadecimal }

  -- Normal rational literals (:: Fractional a => a, from Rational)
  @floating_point		{ strtoken tok_float }
376 377
}

378
<0> {
379 380 381
  -- Unboxed ints (:: Int#)
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
382 383 384 385 386 387
  @decimal \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
  0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
  0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
  @negative @decimal \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
  @negative 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
388 389 390

  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
391 392
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
393 394 395 396 397 398
}

-- Strings and chars are lexed by hand-written code.  The reason is
-- that even if we recognise the string or char here in the regex
-- lexer, we would still have to parse the string afterward in order
-- to convert it to a String.
399
<0> {
400 401 402 403 404 405 406 407
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

{
-- -----------------------------------------------------------------------------
-- The token type

408
data Token
409 410 411 412 413 414
  = ITas  			-- Haskell keywords
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
415
  | ITderive
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
  | ITdo
  | ITelse
  | IThiding
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere
  | ITscc			-- ToDo: remove (we use {-# SCC "..." #-} now)

  | ITforall			-- GHC extension keywords
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
  | ITthreadsafe
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
  | ITdotnet
  | ITmdo
448
  | ITfamily
449

450 451 452 453
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
454 455 456 457 458
  | ITsource_prag
  | ITrules_prag
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
459
  | ITgenerated_prag
460
  | ITcore_prag                 -- hdaume: core annotations
461
  | ITunpack_prag
462
  | ITclose_prag
463 464 465
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528

  | ITdotdot  			-- reserved symbols
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

  | ITbiglam			-- GHC-extension symbols

  | ITocurly  			-- special symbols
  | ITccurly
  | ITocurlybar                 -- {|, for type applications
  | ITccurlybar                 -- |}, for type applications
  | ITvocurly
  | ITvccurly
  | ITobrack
  | ITopabrack			-- [:, for parallel arrays with -fparr
  | ITcpabrack			-- :], for parallel arrays with -fparr
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote

  | ITvarid   FastString	-- identifiers
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)

  | ITdupipvarid   FastString	-- GHC extension: implicit param: ?x

  | ITpragma StringBuffer

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
  | ITrational   Rational

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
529 530 531 532 533 534 535 536 537
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
538 539 540 541

  -- Arrow notation extension
  | ITproc
  | ITrec
542 543 544 545 546 547
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
548 549 550

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
551 552 553 554 555 556 557 558

  -- Documentation annotations
  | ITdocCommentNext  String     -- something beginning '-- |'
  | ITdocCommentPrev  String     -- something beginning '-- ^'
  | ITdocCommentNamed String     -- something beginning '-- $'
  | ITdocSection      Int String -- a section heading
  | ITdocOptions      String     -- doc options (prune, ignore-exports, etc)

559 560 561 562
#ifdef DEBUG
  deriving Show -- debugging
#endif

563
isSpecial :: Token -> Bool
564 565 566 567 568
-- If we see M.x, where x is a keyword, but
-- is special, we treat is as just plain M.x, 
-- not as a keyword.
isSpecial ITas        	= True
isSpecial IThiding    	= True
569
isSpecial ITderive    	= True
570 571 572 573 574 575 576 577 578 579 580
isSpecial ITqualified 	= True
isSpecial ITforall    	= True
isSpecial ITexport    	= True
isSpecial ITlabel     	= True
isSpecial ITdynamic   	= True
isSpecial ITsafe    	= True
isSpecial ITthreadsafe 	= True
isSpecial ITunsafe    	= True
isSpecial ITccallconv   = True
isSpecial ITstdcallconv = True
isSpecial ITmdo		= True
581
isSpecial ITfamily	= True
582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
isSpecial _             = False

-- the bitmap provided as the third component indicates whether the
-- corresponding extension keyword is valid under the extension options
-- provided to the compiler; if the extension corresponding to *any* of the
-- bits set in the bitmap is enabled, the keyword is valid (this setup
-- facilitates using a keyword in two different extensions that can be
-- activated independently)
--
reservedWordsFM = listToUFM $
	map (\(x, y, z) -> (mkFastString x, (y, z)))
       [( "_",		ITunderscore, 	0 ),
	( "as",		ITas, 		0 ),
	( "case",	ITcase, 	0 ),     
	( "class",	ITclass, 	0 ),    
	( "data",	ITdata, 	0 ),     
	( "default",	ITdefault, 	0 ),  
	( "deriving",	ITderiving, 	0 ), 
600
	( "derive",	ITderive, 	0 ), 
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
	( "do",		ITdo, 		0 ),       
	( "else",	ITelse, 	0 ),     
	( "hiding",	IThiding, 	0 ),
	( "if",		ITif, 		0 ),       
	( "import",	ITimport, 	0 ),   
	( "in",		ITin, 		0 ),       
	( "infix",	ITinfix, 	0 ),    
	( "infixl",	ITinfixl, 	0 ),   
	( "infixr",	ITinfixr, 	0 ),   
	( "instance",	ITinstance, 	0 ), 
	( "let",	ITlet, 		0 ),      
	( "module",	ITmodule, 	0 ),   
	( "newtype",	ITnewtype, 	0 ),  
	( "of",		ITof, 		0 ),       
	( "qualified",	ITqualified, 	0 ),
	( "then",	ITthen, 	0 ),     
	( "type",	ITtype, 	0 ),     
	( "where",	ITwhere, 	0 ),
	( "_scc_",	ITscc, 		0 ),		-- ToDo: remove

Ian Lynagh's avatar
Ian Lynagh committed
621
      	( "forall",	ITforall,	 bit explicitForallBit),
Ian Lynagh's avatar
Ian Lynagh committed
622
	( "mdo",	ITmdo,		 bit recursiveDoBit),
623
	( "family",	ITfamily,	 bit tyFamBit),
624 625 626 627 628 629 630 631 632 633 634 635 636

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
	( "dotnet",     ITdotnet,	 bit ffiBit),

	( "rec",	ITrec,		 bit arrowsBit),
637
	( "proc",	ITproc,		 bit arrowsBit)
638 639
     ]

Ian Lynagh's avatar
Ian Lynagh committed
640
reservedSymsFM :: UniqFM (Token, Int -> Bool)
641
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
    map (\ (x,y,z) -> (mkFastString x,(y,z)))
      [ ("..",  ITdotdot,   always)
        -- (:) is a reserved op, meaning only list cons
       ,(":",   ITcolon,    always)
       ,("::",  ITdcolon,   always)
       ,("=",   ITequal,    always)
       ,("\\",  ITlam,      always)
       ,("|",   ITvbar,     always)
       ,("<-",  ITlarrow,   always)
       ,("->",  ITrarrow,   always)
       ,("@",   ITat,       always)
       ,("~",   ITtilde,    always)
       ,("=>",  ITdarrow,   always)
       ,("-",   ITminus,    always)
       ,("!",   ITbang,     always)

        -- For data T (a::*) = MkT
659
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
660
        -- For 'forall a . t'
Ian Lynagh's avatar
Ian Lynagh committed
661
       ,(".", ITdot, explicitForallEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
662 663 664 665 666

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
667 668

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
669 670
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
671 672
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
673 674 675
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
676 677 678
        -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
        -- form part of a large operator.  This would let us have a better
        -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
679
#endif
680 681 682 683 684
       ]

-- -----------------------------------------------------------------------------
-- Lexer actions

685
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
686

687 688
special :: Token -> Action
special tok span _buf len = return (L span tok)
689

690 691 692
token, layout_token :: Token -> Action
token t span buf len = return (L span t)
layout_token t span buf len = pushLexState layout >> return (L span t)
693

694 695
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
696

697 698 699
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
700

701 702 703
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
704

705
init_strtoken :: Int -> (String -> Token) -> Action
706
-- like strtoken, but drops the last N character(s)
707 708
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
709 710

begin :: Int -> Action
711
begin code _span _str _len = do pushLexState code; lexToken
712 713

pop :: Action
714
pop _span _buf _len = do popLexState; lexToken
715 716

pop_and :: Action -> Action
717
pop_and act span buf len = do popLexState; act span buf len
718

719 720 721 722 723
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
724

725
notFollowedBySymbol _ _ _ (AI _ _ buf)
726
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
727

728 729 730 731 732
-- We must reject doc comments as being ordinary comments everywhere.
-- In some cases the doc comment will be selected as the lexeme due to
-- maximal munch, but not always, because the nested comment rule is
-- valid in all states, but the doc-comment rules are only valid in
-- the non-layout states.
733
isNormalComment bits _ _ (AI _ _ buf)
734 735 736 737 738 739 740
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
741 742 743 744

haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)

745
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
746

747 748
ifExtension pred bits _ _ _ = pred bits

749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
    worker commentAcc input docType oneLine = case alexGetChar input of
      Just ('\n', input') 
        | oneLine -> docCommentEnd input commentAcc docType buf span
        | otherwise -> case checkIfCommentLine input' of
          Just input -> worker ('\n':commentAcc) input docType False
          Nothing -> docCommentEnd input commentAcc docType buf span
      Just (c, input) -> worker (c:commentAcc) input docType oneLine
      Nothing -> docCommentEnd input commentAcc docType buf span
      
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
        check input = case alexGetChar input of
          Just ('-', input) -> case alexGetChar input of
            Just ('-', input) -> case alexGetChar input of
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

        dropNonNewlineSpace input = case alexGetChar input of
          Just (c, input') 
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

777 778 779 780
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
781 782
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
783 784
  input <- getInput
  go 1 input
785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818
  where
    go 0 input = do setInput input; cont
    go n input = case alexGetChar input of
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('\125',input) -> go (n-1) input
        Just (c,_)          -> go n input
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> go (n+1) input
        Just (c,_)       -> go n input
      Just (c,input) -> go n input

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
    go commentAcc input docType _ = case alexGetChar input of
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing -> errBrace input span
        Just ('\125',input@(AI end _ buf2)) ->
          docCommentEnd input commentAcc docType buf span
        Just (c,_) -> go ('-':commentAcc) input docType False
      Just ('\123', input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> do
          setInput input
          let cont = do input <- getInput; go commentAcc input docType False
          nested_comment cont span buf _len
        Just (c,_) -> go ('\123':commentAcc) input docType False
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
819 820 821 822 823 824
  input@(AI _ _ buf) <- getInput
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
    '*' -> lexDocSection 1 input 
825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
      Just (c, _) -> lexDocComment input (ITdocSection n) True
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

-- docCommentEnd
-------------------------------------------------------------------------------
-- This function is quite tricky. We can't just return a new token, we also
-- need to update the state of the parser. Why? Because the token is longer
-- than what was lexed by Alex, and the lexToken function doesn't know this, so 
-- it writes the wrong token length to the parser state. This function is
-- called afterwards, so it can just update the state. 

-- This is complicated by the fact that Haddock tokens can span multiple lines, 
-- which is something that the original lexer didn't account for. 
-- I have added last_line_len in the parser state which represents the length 
-- of the part of the token that is on the last line. It is now used for layout 
-- calculation in pushCurrentContext instead of last_len. last_len is, like it 
-- was before, the full length of the token, and it is now only used for error
-- messages. /Waern 

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
                 SrcSpan -> P (Located Token) 
docCommentEnd input commentAcc docType buf span = do
  setInput input
  let (AI loc last_offs nextBuf) = input
      comment = reverse commentAcc
      span' = mkSrcSpan (srcSpanStart span) loc
      last_len = byteDiff buf nextBuf
      
      last_line_len = if (last_offs - last_len < 0) 
        then last_offs
        else last_len  
  
  span `seq` setLastToken span' last_len last_line_len
  return (L span' (docType comment))
 
errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
 
865
open_brace, close_brace :: Action
866
open_brace span _str _len = do 
867 868
  ctx <- getContext
  setContext (NoLayout:ctx)
869 870
  return (L span ITocurly)
close_brace span _str _len = do 
871
  popContext
872
  return (L span ITccurly)
873 874 875 876

-- We have to be careful not to count M.<varid> as a qualified name
-- when <varid> is a keyword.  We hack around this by catching 
-- the offending tokens afterward, and re-lexing in a different state.
877
check_qvarid span buf len = do
878 879 880 881 882 883 884 885 886 887 888 889
  case lookupUFM reservedWordsFM var of
	Just (keyword,exts)
	  | not (isSpecial keyword) ->
	  if exts == 0 
	     then try_again
	     else do
		b <- extension (\i -> exts .&. i /= 0)
		if b then try_again
		     else return token
	_other -> return token
  where
	(mod,var) = splitQualName buf len
890
	token     = L span (ITqvarid (mod,var))
891 892

	try_again = do
893 894
		(AI _ offs _) <- getInput	
		setInput (AI (srcSpanStart span) (offs-len) buf)
895 896 897 898 899 900 901 902 903 904
		pushLexState bad_qvarid
		lexToken

qvarid buf len = ITqvarid $! splitQualName buf len
qconid buf len = ITqconid $! splitQualName buf len

splitQualName :: StringBuffer -> Int -> (FastString,FastString)
-- takes a StringBuffer and a length, and returns the module name
-- and identifier parts of a qualified name.  Splits at the *last* dot,
-- because of hierarchical module names.
905
splitQualName orig_buf len = split orig_buf orig_buf
906
  where
907 908 909 910 911 912
    split buf dot_buf
	| orig_buf `byteDiff` buf >= len  = done dot_buf
	| c == '.'                	  = found_dot buf'
	| otherwise               	  = split buf' dot_buf
      where
       (c,buf') = nextChar buf
913 914 915 916
  
    -- careful, we might get names like M....
    -- so, if the character after the dot is not upper-case, this is
    -- the end of the qualifier part.
917 918 919 920 921 922 923 924 925 926 927
    found_dot buf -- buf points after the '.'
	| isUpper c    = split buf' buf
	| otherwise    = done buf
      where
       (c,buf') = nextChar buf

    done dot_buf =
	(lexemeToFastString orig_buf (qual_size - 1),
	 lexemeToFastString dot_buf (len - qual_size))
      where
	qual_size = orig_buf `byteDiff` dot_buf
928

929
varid span buf len = 
930 931 932
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
933
		return (L span keyword)
934 935 936
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
937 938 939
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
940 941 942 943 944 945 946 947 948 949 950 951
  where
	fs = lexemeToFastString buf len

conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

qvarsym buf len = ITqvarsym $! splitQualName buf len
qconsym buf len = ITqconsym $! splitQualName buf len

varsym = sym ITvarsym
consym = sym ITconsym

952
sym con span buf len = 
953 954
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
955
		b <- extension exts
956 957 958
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
959 960 961
  where
	fs = lexemeToFastString buf len

962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981
-- Variations on the integral numeric literal.
tok_integral :: (Integer -> Token)
     -> (Integer -> Integer)
 --    -> (StringBuffer -> StringBuffer) -> (Int -> Int)
     -> Int -> Int
     -> (Integer, (Char->Int)) -> Action
tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
  return $ L span $ itint $! transint $ parseUnsignedInteger
     (offsetBytes transbuf buf) (subtract translen len) radix char_to_int

-- some conveniences for use with tok_integral
tok_num = tok_integral ITinteger
tok_primint = tok_integral ITprimint
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
982
tok_float        str = ITrational   $! readRational str
983 984
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
985 986 987 988 989 990

-- -----------------------------------------------------------------------------
-- Layout processing

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
991
do_bol span _str _len = do
992
	pos <- getOffside
993 994 995 996 997
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
998
		return (L span ITvccurly)
999 1000 1001
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1002
		return (L span ITsemi)
1003 1004 1005 1006 1007 1008 1009
	    GT -> do
		popLexState
		lexToken

-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
maybe_layout ITdo	= pushLexState layout_do
1010
maybe_layout ITmdo	= pushLexState layout_do
1011 1012 1013
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1014
maybe_layout ITrec	= pushLexState layout
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
maybe_layout _	        = return ()

-- Pushing a new implicit layout context.  If the indentation of the
-- next token is not greater than the previous layout context, then
-- Haskell 98 says that the new layout context should be empty; that is
-- the lexer must generate {}.
--
-- We are slightly more lenient than this: when the new context is started
-- by a 'do', then we allow the new context to be at the same indentation as
-- the previous context.  This is what the 'strict' argument is for.
--
1026
new_layout_context strict span _buf _len = do
1027
    popLexState
1028
    (AI _ offset _) <- getInput
1029 1030 1031 1032 1033 1034 1035 1036
    ctx <- getContext
    case ctx of
	Layout prev_off : _  | 
	   (strict     && prev_off >= offset  ||
	    not strict && prev_off > offset) -> do
		-- token is indented to the left of the previous context.
		-- we must generate a {} sequence now.
		pushLexState layout_left
1037
		return (L span ITvocurly)
1038 1039
	other -> do
		setContext (Layout offset : ctx)
1040
		return (L span ITvocurly)
1041

1042
do_layout_left span _buf _len = do
1043 1044
    popLexState
    pushLexState bol  -- we must be at the start of a line
1045
    return (L span ITvccurly)
1046 1047 1048 1049

-- -----------------------------------------------------------------------------
-- LINE pragmas

1050 1051
setLine :: Int -> Action
setLine code span buf len = do
1052
  let line = parseUnsignedInteger buf len 10 octDecDigit
1053
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1054 1055 1056 1057 1058
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1059 1060
setFile :: Int -> Action
setFile code span buf len = do
1061
  let file = lexemeToFastString (stepOn buf) (len-2)
1062
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1063 1064 1065 1066
  popLexState
  pushLexState code
  lexToken

1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092

-- -----------------------------------------------------------------------------
-- Options, includes and language pragmas.

lex_string_prag :: (String -> Token) -> Action
lex_string_prag mkTok span buf len
    = do input <- getInput
         start <- getSrcLoc
         tok <- go [] input
         end <- getSrcLoc
         return (L (mkSrcSpan start end) tok)
    where go acc input
              = if isString input "#-}"
                   then do setInput input
                           return (mkTok (reverse acc))
                   else case alexGetChar input of
                          Just (c,i) -> go (c:acc) i
                          Nothing -> err input
          isString i [] = True
          isString i (x:xs)
              = case alexGetChar i of
                  Just (c,i') | c == x    -> isString i' xs
                  _other -> False
          err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"


1093 1094 1095 1096 1097 1098
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
1099
lex_string_tok span buf len = do
1100 1101
  tok <- lex_string ""
  end <- getSrcLoc 
1102
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1103

1104
lex_string :: String -> P Token
1105 1106
lex_string s = do
  i <- getInput
1107
  case alexGetChar' i of
1108 1109 1110 1111
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1112 1113
	magicHash <- extension magicHashEnabled
	if magicHash
1114 1115
	  then do
	    i <- getInput
1116
	    case alexGetChar' i of
1117 1118 1119 1120
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1121
                    else let s' = mkZFastString (reverse s) in
1122
			 return (ITprimstring s')
1123 1124
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
	      _other ->
		return (ITstring (mkFastString (reverse s)))
	  else
		return (ITstring (mkFastString (reverse s)))

    Just ('\\',i)
	| Just ('&',i) <- next -> do 
		setInput i; lex_string s
	| Just (c,i) <- next, is_space c -> do 
		setInput i; lex_stringgap s
1135
	where next = alexGetChar' i
1136

1137 1138 1139
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149

lex_stringgap s = do
  c <- getCharOrFail
  case c of
    '\\' -> lex_string s
    c | is_space c -> lex_stringgap s
    _other -> lit_error


lex_char_tok :: Action
1150 1151 1152 1153 1154 1155
-- Here we are basically parsing character literals, such as 'x' or '\n'
-- but, when Template Haskell is on, we additionally spot
-- 'x and ''T, returning ITvarQuote and ITtyQuote respectively, 
-- but WIHTOUT CONSUMING the x or T part  (the parser does that).
-- So we have to do two characters of lookahead: when we see 'x we need to
-- see if there's a trailing quote
1156
lex_char_tok span buf len = do	-- We've seen '
1157
   i1 <- getInput	-- Look ahead to first character
1158
   let loc = srcSpanStart span
1159
   case alexGetChar' i1 of
1160 1161
	Nothing -> lit_error 

1162
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1163 1164 1165
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1166
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1167 1168
		   else lit_error

1169
	Just ('\\', i2@(AI end2 _ _)) -> do 	-- We've seen 'backslash 
1170 1171 1172 1173
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1174
			        else do setInput i2; lit_error 
1175

1176 1177 1178
        Just (c, i2@(AI end2 _ _)) 
		| not (isAny c) -> lit_error
		| otherwise ->
1179 1180 1181

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
1182
	   case alexGetChar' i2 of	-- Look ahead one more character
1183 1184 1185 1186 1187 1188 1189
		Nothing -> lit_error
		Just ('\'', i3) -> do 	-- We've seen 'x'
			setInput i3 
			finish_char_tok loc c
		_other -> do 		-- We've seen 'x not followed by quote
					-- If TH is on, just parse the quote only
			th_exts <- extension thEnabled	
1190 1191
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1192
				   else do setInput i2; lit_error
1193

1194
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1195 1196
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
1197
  = do	magicHash <- extension magicHashEnabled
1198
	i@(AI end _ _) <- getInput
1199
	if magicHash then do
1200
		case alexGetChar' i of
1201
			Just ('#',i@(AI end _ _)) -> do
1202
				setInput i
1203
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1204
			_other ->
1205 1206 1207 1208
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1209 1210 1211 1212 1213
lex_char :: Char -> AlexInput -> P Char
lex_char c inp = do
  case c of
      '\\' -> do setInput inp; lex_escape
      c | isAny c -> do setInput inp; return c
1214
      _other -> lit_error
1215

1216 1217 1218
isAny c | c > '\xff' = isPrint c
	| otherwise  = is_any c

1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237
lex_escape :: P Char
lex_escape = do
  c <- getCharOrFail
  case c of
	'a'   -> return '\a'
	'b'   -> return '\b'
	'f'   -> return '\f'
	'n'   -> return '\n'
	'r'   -> return '\r'
	't'   -> return '\t'
	'v'   -> return '\v'
	'\\'  -> return '\\'
	'"'   -> return '\"'
	'\''  -> return '\''
	'^'   -> do c <- getCharOrFail
		    if c >= '@' && c <= '_'
			then return (chr (ord c - ord '@'))
			else lit_error

1238 1239 1240
	'x'   -> readNum is_hexdigit 16 hexDigit
	'o'   -> readNum is_octdigit  8 octDecDigit
	x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x)
1241 1242 1243

	c1 ->  do
	   i <- getInput
1244
	   case alexGetChar' i of
1245 1246
	    Nothing -> lit_error
	    Just (c2,i2) -> 
1247 1248
              case alexGetChar' i2 of
		Nothing	-> do setInput i2; lit_error
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
		Just (c3,i3) -> 
		   let str = [c1,c2,c3] in
		   case [ (c,rest) | (p,c) <- silly_escape_chars,
			      	     Just rest <- [maybePrefixMatch p str] ] of
			  (escape_char,[]):_ -> do
				setInput i3
				return escape_char
			  (escape_char,_:_):_ -> do
				setInput i2
				return escape_char
			  [] -> lit_error

readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
readNum is_digit base conv = do
1263
  i <- getInput
1264 1265 1266
  c <- getCharOrFail
  if is_digit c 
	then readNum2 is_digit base conv (conv c)
1267
	else do setInput i; lit_error
1268 1269 1270 1271 1272

readNum2 is_digit base conv i = do
  input <- getInput
  read i input
  where read i input = do
1273
	  case alexGetChar' input of
1274 1275 1276 1277
	    Just (c,input') | is_digit c -> do
		read (i*base + conv c) input'
	    _other -> do
		if i >= 0 && i <= 0x10FFFF
1278
		   then do setInput input; return (chr i)
1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
		   else lit_error

silly_escape_chars = [
	("NUL", '\NUL'),
	("SOH", '\SOH'),
	("STX", '\STX'),
	("ETX", '\ETX'),
	("EOT", '\EOT'),
	("ENQ", '\ENQ'),
	("ACK", '\ACK'),
	("BEL", '\BEL'),
	("BS", '\BS'),
	("HT", '\HT'),
	("LF", '\LF'),
	("VT", '\VT'),
	("FF", '\FF'),
	("CR", '\CR'),
	("SO", '\SO'),
	("SI", '\SI'),
	("DLE", '\DLE'),
	("DC1", '\DC1'),
	("DC2", '\DC2'),
	("DC3", '\DC3'),
	("DC4", '\DC4'),
	("NAK", '\NAK'),
	("SYN", '\SYN'),
	("ETB", '\ETB'),
	("CAN", '\CAN'),
	("EM", '\EM'),
	("SUB", '\SUB'),
	("ESC", '\ESC'),
	("FS", '\FS'),
	("GS", '\GS'),
	("RS", '\RS'),
	("US", '\US'),
	("SP", '\SP'),
	("DEL", '\DEL')
	]

1318 1319 1320 1321
-- before calling lit_error, ensure that the current input is pointing to
-- the position of the error in the buffer.  This is so that we can report
-- a correct location to the user, but also so we can detect UTF-8 decoding
-- errors if they occur.
1322 1323 1324 1325 1326
lit_error = lexError "lexical error in string/character literal"

getCharOrFail :: P Char
getCharOrFail =  do
  i <- getInput
1327
  case alexGetChar' i of
1328 1329 1330
	Nothing -> lexError "unexpected end-of-file in string/character literal"
	Just (c,i)  -> do setInput i; return c

1331 1332 1333 1334 1335 1336 1337 1338
-- -----------------------------------------------------------------------------
-- Warnings

warn :: DynFlag -> SDoc -> Action
warn option warning span _buf _len = do
    addWarning option (mkWarnMsg span alwaysQualify warning)
    lexToken

1339 1340 1341 1342 1343 1344
-- -----------------------------------------------------------------------------
-- The Parse Monad

data LayoutContext
  = NoLayout
  | Layout !Int
1345
  deriving Show
1346 1347 1348 1349

data ParseResult a
  = POk PState a
  | PFailed 
1350
	SrcSpan		-- The start and end of the text span related to
1351 1352
			-- the error.  Might be used in environments which can 
			-- show this span, e.g. by highlighting it.
1353 1354 1355 1356
	Message		-- The error message

data PState = PState { 
	buffer	   :: StringBuffer,
1357 1358
    dflags     :: DynFlags,
    messages   :: Messages,
1359
        last_loc   :: SrcSpan,	-- pos of previous token
1360 1361 1362
        last_offs  :: !Int, 	-- offset of the previous token from the
				-- beginning of  the current line.
				-- \t is equal to 8 spaces.
1363
	last_len   :: !Int,	-- len of previous token
1364
  last_line_len :: !Int,
1365 1366 1367 1368 1369
        loc        :: SrcLoc,   -- current loc (end of prev token + 1)
	extsBitmap :: !Int,	-- bitmap that determines permitted extensions
	context	   :: [LayoutContext],
	lex_state  :: [Int]
     }
1370
	-- last_loc and last_len are used when generating error messages,
1371 1372 1373 1374